The moment freelance journalist Caroline Williams learned she was being impersonated by a suspected AI scammer, she feared her livelihood was under threat.
She had just discovered that Science, one of the world’s top peer-reviewed journals, had published an article under her name.
The article was on her area of expertise, about which she has written numerous pieces, and a book.
But the new Science article had not been written by Caroline.
It had been pitched to editors by someone else. Not only did it contain plagiarised quotes from other articles, it was also produced using AI, detection software suggested.
There are widespread concerns about the potential for artificial intelligence to make a swathe of jobs redundant. But the technology also makes it much easier and cheaper for criminals to commit fraud, and on a bigger scale – with particular concern about impersonation scams.
“I was absolutely horrified,” Caroline tells The i Paper. “I had no control about what was happening and no way to stop it.”
‘Fake Caroline’
Caroline first learned of the suspected impersonation in August. An editor at the American magazine Chemistry and Engineering News said she had been contacted by someone proposing an article, but using a different email address to Caroline’s usual one.
She didn’t think much of it at the time, but soon afterwards, she was contacted by another editor – this time from science journal Nature – who reported the same thing.
Alongside their pitch, the emailer had claimed to be the author of Caroline’s previous newspaper stories. But after exchanging further emails, and getting poorly written replies, the editors became suspicious and alerted her.
She put a warning about the “Fake Caroline” on LinkedIn, in case other editors had been contacted.
“They are sending editors rambling and badly thought through pitches on subjects I care about, pointing editors to my book and my clips,” she wrote. “EDITORS: THIS IS NOT ME!!!”
The successful scam
That’s when editors at Science discovered they had already been scammed by “Fake Caroline” and published an article.
While the piece had needed heavy editing, they had paid the person $1,000 (£750), for a feature article headlined “Your heartbeat quietly shapes how your brain processes information.”
The piece was on Caroline’s expert area: interoception, the way we sense our body’s inner workings, such as our heartbeat. She has also written a book on the topic.
“The editor was surprised at the low quality of the story but decided to try to save it and work with the writer, as we often do,” says John Travis, managing news editor at Science.
Once the deception was discovered, Science used AI-detecting software that they claimed found the first draft had been written using AI. It also contained quotes copied from other online articles on interoception.
The piece was taken off its website.
“I feel really bad for [Caroline], and I wish we hadn’t gotten fooled on her behalf,” Travis adds.
The journal has since changed its policies so that editors will talk by phone or Zoom with any new writers, and will use AI-detecting software.
But Caroline says: “I was embarrassed that anyone thought this came from me. The only thing you’ve got as a freelancer is your reputation and your relationship with editors and scientists – and that was being eroded.”
Writers aren’t the only group whose living is under threat from AI. In the past few years, advances in “generative AI” has led to programs like ChatGPT that can write articles, dispense therapy, write computer code and produce convincing artwork, videos and music.
But to Caroline it seemed especially unjust that someone was using AI to write worse versions of articles she could have written herself, while trashing her reputation in the process.
‘I’m not an impersonator’
The i Paper contacted “Fake Caroline” using the email address given to Science and the other publications.
Someone responded claiming they were in a fact a US journalist who shared the name of Caroline Williams. “I did not impersonate anyone. The piece published by Science was submitted under my own name and is my own work, and I reject the criticisms described in your email,” they said.
However, in a story pitch seen by The i Paper, “Fake Caroline” claimed to be the real journalist.
“I’ve written a lot about interoception and the mind-body connection for general audiences,” the email pitch stated, before linking to a piece the real Caroline had previously had published in New Scientist magazine.
When asked about this, the “Fake Caroline” said: “Including that link was a mistake on my part. It was not meant to suggest that I wrote that New Scientist piece, and it was not central to my pitch.
“I am a different person from the UK journalist, and I did not intend to present her work as mine.”
When asked for other examples of their work, the person said that “there is no need for additional clips” as the Science article “should be sufficient”. They also declined our offer of a Zoom call.
When this claim was put to Travis, he said the person pitching the story “clearly identified themselves as UK Caroline”.
‘This is a big problem for journalism’
Caroline also does not buy it. And she is not the only writer this has happened to. She knows of two other freelancers who say they have been subject to similar impersonation attempts.
More generally, AI-written articles are an increasing problem in journalism, she warns. Not only may they be badly written, but they contain plagiarism and inaccuracies – sometimes termed hallucinations.
For now, Caroline is continuing her fight to stop what she believes is impersonation. “I’ve emailed every editor I can think of at every magazine that I could conceivably think about pitching to,” she says.
“But it’s possible that there are loads of editors out there that have had fake pitches and either thought that it was the real me and I’m rubbish, or spotted that it was AI and ignored it. Who knows?”
But she adds: “It’s not just about me and my reputation – it’s a problem for journalism in general. AI hallucinations passed off as fact can find their way into common ‘knowledge’ despite being total nonsense – when that’s health-related it could be harmful.”